Gübelin to Develop AI Approach to Gemstone Analysis

Lucerne, Switzerland—Gübelin Gem Lab is taking steps to advance gemstone analysis through artificial intelligence.

The lab is partnering with Swiss research and development center CSEM to develop a platform that will automate the process of determining gemstone authenticity and origin, which currently relies heavily on human judgment and analysis.

Gübelin and CSEM submitted a joint project proposal, titled “Gemtelligence—Software Development for the Automated Analysis of Gemstones,” to Innosuisse, a government institute that promotes science-based innovation.

Innosuisse approved the project, and now the Swiss government will grant CHF 600,000 (about $655,000) toward it, Gübelin said.

The lab said it chose CSEM as its partner because of the center’s expertise with complex and diverse sets of data.

The cutting-edge project will include assessing and evaluating the various types of analytical data generally collected in modern gem labs, including traditional gemological data like microscopic and macroscopic descriptors and observations, color parameters, size, optical parameters and fluorescence behavior.

It also will include more modern data, such as FTIR (Fourier-Transform Infrared Spectrometer) and UV-Vis-nIR spectra, Raman spectra, and chemical trace and ultra-trace element composition.

The two companies will develop machine-learning-based algorithms to evaluate standard gem characteristics, using existing sets of data as a base: a catalog of the tens of thousands of client gemstones the lab has tested since the 1970s along with the Gübelin Reference Stone Collection comprising more than 27,000 gems.

Gübelin Managing Director Daniel Nyfeler said the new technology will increase the consistency and reliability of gem lab results, reduce human error and allow for the scalability of gem testing services.

The project is expected to span three years but will yield results that likely will be integrated into the daily work of the lab earlier than that, Gübelin said.

Philipp Schmid, head of Industry 4.0 and Machine Learning at CSEM, said: “We are dealing with data with different degree of structuring, ranging from spectra and chemical element concentrations to microscopy images, handwritten descriptions and subjective reflections by various experts. The goal is to create a kind of super-expert, working hand-in-hand with the human experts.”

Gübelin said it started using automated data evaluation about 10 years ago, mostly for large sets of chemical data and to harmonize gem interpretations throughout its labs.

The use of AI in gemology has largely been a support tool up to now, but that is changing.

In addition to Gübelin’s project, the Gemological Institute of America also is